Conventional object detection applications extract a significant number of multi-scale features from an image in order to enable detection of an object in the image. For example in face detection, conventional solutions may extract approximately 200,000 features in an image with a maximum dimension of 1480. It is to be appreciated that extracting such a high number of features can be very time consuming, and thus becomes the bottleneck of conventional object detection methods in terms of speed.